Skip to content

guidellm

Guidellm is a package that provides an easy and intuitive interface for evaluating and benchmarking large language models (LLMs).

configure_logger(config=settings.logging)

Configure the metrics for LLM Compressor. This function sets up the console and file logging as per the specified or default parameters.

Note: Environment variables take precedence over the function parameters.

Parameters:

Name Type Description Default
config LoggingSettings

The configuration for the logger to use.

logging
Source code in src/guidellm/logger.py
def configure_logger(config: LoggingSettings = settings.logging):
    """
    Configure the metrics for LLM Compressor.
    This function sets up the console and file logging
    as per the specified or default parameters.

    Note: Environment variables take precedence over the function parameters.

    :param config: The configuration for the logger to use.
    :type config: LoggerConfig
    """

    if config.disabled:
        logger.disable("guidellm")
        return

    logger.enable("guidellm")

    if config.clear_loggers:
        logger.remove()

    # log as a human readable string with the time, function, level, and message
    logger.add(
        sys.stdout,
        level=config.console_log_level.upper(),
        format="<green>{time:YY-MM-DD HH:mm:ss}</green>|<level>{level: <8}</level> \
        |<cyan>{name}:{function}</cyan>:<cyan>{line}</cyan> - <level>{message}</level>",
    )

    if config.log_file or config.log_file_level:
        log_file = config.log_file or "guidellm.log"
        log_file_level = config.log_file_level or "INFO"
        # log as json to the file for easier parsing
        logger.add(log_file, level=log_file_level.upper(), serialize=True)

reload_settings()

Reload the settings from the environment variables

Source code in src/guidellm/settings.py
def reload_settings():
    """
    Reload the settings from the environment variables
    """
    new_settings = Settings()
    settings.__dict__.update(new_settings.__dict__)